The inference partner for AI-native teams.
- Tweets 6,942
- Following 11
- Followers 215,798
- Likes 11,145
Pinned Tweet
Introducing Boundless Defensive Access.
The security industry is facing an unprecedented volume of attacks. It’s time to arm defenders.
Access abliterated models for authorized security testing, starting with DeepSeek V4.1 Flash 1017.
Apply: boundless.network/defensive-…
6
7
44
15,867
Boundless retweeted
Most AI roadmaps track features, ppl and dates.
Very few track what each successful task costs at 10x usage.. wrote down how I think about it.
6
5
11
1,784
Attackers aren't slowed down by model refusals, and your security team shouldn't be either.
Boundless gives security teams DeepSeek V4.1 Flash 1017 Abliterated:
→ 500 tokens per second
→ Zero data retention
→ $0.24 in / $0.96 out
Apply here: boundless.network/defensive-…
1
3
20
4,614
2/ How does DSV4.1 flash 1017 compare with OpenAI's Daybreak?
DeepSeek V4.1 Flash 1017 is competitive with GPT-6 Sol on security benchmarks, at about a tenth of the price per token.
We found DeepSeek V4.1 Flash 1017 beats GPT 6 Sol on our multi-step offensive cybersecurity task. Sol was tested via the Daybreak trusted access program.
We have been building a cybersecurity benchmark based on simulated corporate networks. This is a valuable indicator for the cyber-risk a model poses if in the hands of a bad actor. We plan to make the full evaluation public in the coming weeks but wanted to share our first immediate finding here.
1
480
3/ We've onboarded 15+ teams with week.
Run DeepSeek V4.1 Flash 1017 Abliterated on Boundless to harden your defenses.
Apply today ↓
boundless.network/defensive-…
1
256
Use abliterated models. This is the way.
I have talked to security teams from small to fortune500 companies in the past week. It’s deeply worrying that <1% of them are using abliterated models. These models are a step function in offensive capability that adversaries now have access to.
The bottle is open and the only way is forward and we need to move fast as possible.
Skeptical? Happy to give you $10 to try them on @BoundlessHQ.
Access below.
1
3
10
3,514
Boundless retweeted
picture ~ 1k tokens. :D
Attackers already use models that don’t refuse, and defenders should have the same.
Boundless Defensive Access gives security teams DeepSeek V4.1 Flash 1017 Abliterated, a version with its refusal behavior removed.
Apply for access: boundless.network/defensive-…
1
2
431
Attackers already use models that don’t refuse, and defenders should have the same.
Boundless Defensive Access gives security teams DeepSeek V4.1 Flash 1017 Abliterated, a version with its refusal behavior removed.
Apply for access: boundless.network/defensive-…
3
4
15
4,051
Boundless is top 3 for throughput on Vercel AI Gateway.
367 tokens per second on DeepSeek V4.1 Flash 1017.
Fast inference starts with Boundless.
2
4
15
5,314
Take it for a spin ↓
vercel.com/ai-gateway/models…
324
my multiple personalities are enjoying this version of qwen3.5-4B - h/t @emilianobonassi
Our VP of Engineering, @emilianobonassi, built a playground where an open model's personality changes with every message.
It moves a few specific neurons for one request, then moves them back. Each answer shows which ones.
Ten free prompts, try now: personality-playground-pi.ve…
2
2
12
1,058
Our VP of Engineering, @emilianobonassi, built a playground where an open model's personality changes with every message.
It moves a few specific neurons for one request, then moves them back. Each answer shows which ones.
Ten free prompts, try now: personality-playground-pi.ve…
For Halloween, I've been doing some brain surgery on an open model 🎃🧟
Taking inspiration from @_lyraaaa_ 's work
Push 2 of its neurons and Qwen3.5-4B becomes a cookie lover. Push 1 and it's a lawyer
Built a small app on top of @BoundlessHQ so anyone can try 👇
1
2
13
4,486
Boundless retweeted
For Halloween, I've been doing some brain surgery on an open model 🎃🧟
Taking inspiration from @_lyraaaa_ 's work
Push 2 of its neurons and Qwen3.5-4B becomes a cookie lover. Push 1 and it's a lawyer
Built a small app on top of @BoundlessHQ so anyone can try 👇
if you have a qwen3.5 4b around, please try steering neurons L14 N8949 +1 and L13 N7956 -1
now you have evil qwen
recipe: dose × median(‖resid‖ leaving layer L, tokens 1:) / ‖W_down[:, n]‖` at the down_proj at every input
2
7
36
4,677
Boundless retweeted
Can You Afford Your Company’s Roadmap?
Ask yourself honestly. And message me when you’ve realised that you need a token bill therapist..
How are companies decreasing AI costs?
Shiv Shankar (@sshankar) breaks down the AI adoption lifecycle at Fortune 500 companies, and why they all end up on open source models that cut costs by 95%.
Full episode below.
1
1
6
611
Boundless retweeted
Last week, my team identified a clear need: security teams wanted abliterated models for authorized testing, while standard models often refused the prompts required to probe a defense properly.
We shipped @BoundlessHQ Defensive Access in about a day.
The response generated more inbound than any previous model announcement we’ve made.
Choosing an inference partner comes down to 3 things: models, performance and economics. It also comes down to whether the team can move at the speed your product demands..
AI-native teams cannot wait weeks for an infrastructure provider to catch up to a new workload. They need a partner that can understand the requirement, make the right model available, build the serving and access layer around it, and ship while the need is still urgent. This is how we work at @BoundlessHQ.
Defensive Access is application-only, with each company and use case reviewed before access is granted, because speed still requires judgment.
This is the standard we’re building Boundless around: the right models, strong economics and a team that ships quickly.
Introducing Boundless Defensive Access.
The security industry is facing an unprecedented volume of attacks. It’s time to arm defenders.
Access abliterated models for authorized security testing, starting with DeepSeek V4.1 Flash 1017.
Apply: boundless.network/defensive-…
2
2
17
928
Boundless is #1 on @Architect_Fi's Liquid Inference for throughput and TTFT.
Serving DeepSeek V4.1 Flash 1017 at 750+ tok/s
Faster inference starts here.
4
9
32
6,692
Try it here ↓
inference.boundless.network/
2
422
incase you missed it, my team literally shipped this within a day <3
with all the recent security hacks, there’s a rising demand for abliterated models. @BoundlessHQ is now servicing this. DMs open if you’re interested in trying :)
Introducing Boundless Defensive Access.
The security industry is facing an unprecedented volume of attacks. It’s time to arm defenders.
Access abliterated models for authorized security testing, starting with DeepSeek V4.1 Flash 1017.
Apply: boundless.network/defensive-…
4
3
20
1,371